Relationships vitally support individuals' physical and mental health; social isolation, for instance, constitutes a powerful risk factor for mortality, cardiovascular disease, and mood disorders. The predominant scientific model of this phenomenon holds that social ties bolster health through stress buffering: reduced stress reactivity in the presence of supportive others. Although stress buffering offers considerable explanatory power, not all relationships reduce individuals' stress; some forms of support fail to reduce-and can even worsen-subjective and physiological indices of stress. Scientists have begun to examine the specific features of relationships and support behaviors that may moderate the effects of relationships on stress and mental health, but these efforts have been limited in at least two ways. First, this work focuses almost exclusively on relationship-level predictors of stress buffering (e.g., marital satisfaction) as opposed to characteristics of individuals who effectively buffer others' stress. This is especially important because research on empathy suggests that individuals who insightfully understand others' affective states (empathic accuracy) and tend to vicariously share those states (affect sharing) might also engage in prosocial and adaptive interpersonal behaviors. Second, existing work has tended to examine stress buffering using naturalistic self-report measures (e.g., daily diaries) or controlle laboratory tasks, but has rarely combined these techniques, preventing an integration of knowledge garnered by each approach. The proposed work will address these gaps in knowledge and propose a multilevel, integrative, and conceptually novel model of stress buffering. This model posits that empathic individuals provide high quality support that, in turn, reduces support recipients' stress and negative affect. We will test this model using a hybrid laboratory and field paradigm our group has recently developed. We will select pairs of close friends and examine empathy in one member of each pair through a behavioral marker of empathic accuracy-performance on an accuracy task we have developed-and a neural marker of affect sharing we have also developed-individuals' engagement of mesolimbic dopaminergic targets while watching their friend receive monetary prizes. Friend pairs will then complete daily diaries reporting on their patterns of stress and social support. Finally, we will collect samples f diurnal salivary cortisol-a canonical neuroendocrine measure of stress reactivity-and reports of sustained threat and loss (subdomains of the Negative Valence Systems RDoC domain), from the other member of the friend pair. We predict that neural and behavioral markers of empathy in support providers will predict support recipients' reductions in endocrine responses to stress, as well as reductions in subjective sustained threat and loss, and that this relationship will be mediated by the quality of support high empathy individuals provide. These data will provide multiple novel insights concerning the social underpinnings of psychological well-being and pave the way for translational work aimed at improving mental health on a broad scale.